Swine flu: what are the impacts on the New Zealand economy - a macro-modelling approach
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Swine flu: what are the impacts on the New
Zealand economy — a macro-modelling
approach
Martin Fukac̆ and Kirdan Lees
Economics Department, RBNZ, P.O.Box 2498, Wellington, New Zealand
Abstract
We adopt a macro-economic modelling approach to estimating the potential im-
pact of swine flu on the New Zealand economy. In particular, we use consumption
and labour shocks to mimic the initial impact of the virus and use the Reserve
Bank’s new KITT model to model the dynamics of the shock transmission. We
find relatively small impacts on the macroeconomy. Based on Ministry of Health
assumptions that suggest less than 200 swine flu related deaths, we find declines in
output of at most 0.62 percent.
1 Introduction
This note considers the impact of swine flu (that is, influenza A H1N1) on the
New Zealand macroeconomy. The approach is similar to the macroeconomic
modelling approach Keogh-Brown et al. (2009) use to model the impact of
an influenza outbreak on the United Kingdom. Our analysis is based around
three sets of assumptions about swine flu: (i) a set of epidemiology assumptions
regarding the clinical attack rate and clinical fatality rate; (ii) an assumption
that predominantly consumption and labour shocks can be used to mimic the
initial macroeconomic impact; and (iii) the use of the KITT model to map
the transmission of the shocks to the rest of the macroeconomy. The results
are conditional on all three sets of assumptions.
Our preferred baseline case is predicated on recent Ministry of Health as-
sumptions that suggest less than 200 swine flu deaths. We find relatively mild
declines in output of, at most, 0.62 percent cumulated over the first year after
the outbreak. The output declines are driven by reduced consumption demand
and reduced supply of labour. These shocks tend to have offsetting implica-
tions for inflation and hence monetary policy. Further, firms and households
13 July 2009may engage in behaviour that smooths or minimizes the impact of the shocks
(such as affected firms meeting demand by reducing stocks or workers deferring
training to cover absent colleagues). These effects are not directly captured in
the KITT macroeconomic model and thus we suggest our estimate of a 0.62
percent fall in output represents an upper threshold on the macroeconomic
impact of swine flu.
Our baseline numbers are smaller than some reported recently in the mar-
kets (see O’Donovan and Zollner, 2009, for example), and elsewhere (the New
Zealand Treasury working paper Douglas et al., 2006, for example). This is
largely a function of using a more benign set of epidemiology numbers that
we associate with no material school closures or prophylactic absenteeism.
In addition to our baseline case, we model a conservative scenario and a severe
scenario associated with a less benign set of epidemiological assumptions and
more disruption on household behaviour. These cases have progressively larger
implications for output, mostly as a function of increased school closures and
absenteeism rather than as a direct function of the number of deaths.
2 The swine flu shock
2.1 Epidemiology
Underlying the macroeconomic impact of the shock are assumptions regarding
the epidemiology or virulence of swine flu. Swine flu is closely related to avian
flu within the influenza family of viruses. While swine flu is a new strain
of influenza, some basis for the proportion of the population who contract
swine flu can be calibrated based on previous outbreaks, and by examining
the pattern of spread of existing cases.
We assume that 30 percent of the population contract swine flu and develop
symptoms. Ministry of Health planning suggests a higher clinical attack rate
(CAR) but estimate just over one one-third of the New Zealand population
will become symptomatic.
Our attack rate is the same attack rate as some earlier studies of influenza
(see Douglas et al., 2006, for example). For comparison, Keogh-Brown et al.
(2009) suggest a clinical attack rate of 25 percent for the United Kingdom
experience of the 1918 influenza epidemic and note higher attack rates for
the 1957 (greater than 30 percent) and 1968/69 epidemics (greater than 45
percent). For the case of the US, James and Sargent (2007) suggest an attack
rate of 25 percent for the 1918 outbreak and 35 percent for the 1957 outbreak.
2However, the clinical attack rate is likely to differ from country to country and
from episode to episode.
The case fatality rate (CFR) also differs widely across influenza outbreaks. We
use a fatality rate recently suggested by the Ministry of Health and consistent
with less than 200 deaths in the New Zealand population. Relative to historical
outbreaks and indeed, the seasonal flu, this is a relatively low fatality rate. We
also explore more severe scenarios associated with a materially different set of
epidemiology assumptions, where the fatality rate is 0.08 percent of those that
contract the illness. Previous influenza outbreaks have often been restricted to
one or two months (for example, deaths due to the 1918 US epidemic spiked
in a single month) and we base our case to the third quarter of 2009. 1
2.2 Impact on the labour force
The impact of the swine flu shock on the labour force is modelled through four
channels (i) deaths from the virus itself; (ii) sickness from contracting swine
flu; (iii) school closures; and (iv) “prophylactic absenteeism”, where either
employees absent themselves from work or employers direct employees not to
attend work, in order not to catch the virus. We approximate the New Zealand
labour force as equivalent to 3 million people and centre our analysis on the
13 weeks available in the quarter that the virus is assumed to hit the economy.
Fatalities from the virus can be calculated by simply multiplying the case
attack rate by the case fatality rate. Here, we take the latest Ministry of
Health release that suggests 200 influenza deaths for the entire population in
New Zealand, which generates a comparatively low fatality rate in the region
of 0.02 percent. Table 1 shows the impact on the New Zealand labour force
(assumed to be approximately 3 million) under the baseline scenario.
We make a number of specific assumptions that are relevant to the labour
force. First we assume that the possibility of death can occur at anytime in
the quarter so the number of weeks out of the labour force due to death
averages 6.5 weeks out of the possible 13 weeks. The direct impact of deaths
on the labour force is insignificant.
Contracting swine flu has a larger impact on the labour force than the number
of deaths from the illness. Both our baseline and conservative case assume the
period of illness associated with catching swine flu is 4-6 days. We assume
that this translates to a period of approximately one week not working. This
generates a decline of 897,600 weeks in aggregate or a 2.3 percent decline
1 Some outbreaks have been characterized by waves of infection. However, we do
not model the incidence of waves of attack in this study.
3in the labour force under the baseline case. Table 1 shows our assumptions
regarding the impact of the swine flu shock on the labour force for the baseline,
conservative and severe scenario. We ignore the impact of deaths on the labour
force since this effect is extremely small.
Both our conservative and severe scenarios assume a less benign form of swine
flu than the baseline case. Our conservative case assumes material school clo-
sures of every school closing for one week. We assume that this is associated
with a 15 percent decline in the labour force (for one week) since parents have
to look after school children. 2 Further, our severe case assumes an additional
impact from prophylactic absenteeism, where 20 percent of the work force oth-
erwise unaffected by swine flu is told not to come to work, in order to avoid
catching swine flu. School closures and prophylactic absenteeism generate just
over half of the labour force shock in the severe scenario. Assumptions about
the magnitude of the aggregate labour force shock are clearly conditional on
assumptions for school closures and prophylactic absenteeism. 3
2.3 Impact on consumption
To examine the impact on consumption, we split consumers into three types:
(i) those hit by the flu (labelled ‘sick’); (ii) those that stay home because
of absenteeism or school closures (labelled ‘Absent, school closure’); and (iii)
normal consumers (those that don’t contract swine flu, and are unaffected by
school closures or prophylactic absenteeism). We then take the New Zealand
consumption bundle and make specific assumptions about expenditure by each
consumer type across types of goods. These expenditures are then weighted
by the proportion of each consumer in each category and the duration of time
each consumer in each category (no more than one week of the thirteen in the
quarter), to produce an aggregate shock for the quarter.
Table 2 shows the impact of our assumptions regarding consumption expen-
diture for the baseline, conservative and severe scenarios. We assume that
illness substantially reduces expenditure for many types of goods. However,
this is temporary and the rightmost three columns of table 3 show assump-
tions regarding how much consumption of each good is delayed rather than
lost permanently. This assumption produces a boost to consumption in the
quarter immediately following the impact of the flu outbreak.
Note that the consumption reductions are more pronounced for the conserva-
2 The 15 percent assumption is identical to the assumption in Keogh-Brown et al.
(2009).
3 The number of reported deaths and the nature of media coverage might impact
on the extent of these effects.
4Table 1
Impact on labour force
Cause Population Individual Impact on Supply
Affected weeks total weeks impact
Baseline Scenario
i) Contracting flu 30.00% 1 897,600 -2.30%
ii) School closures 0.00% 0 0,000 0.00%
iii) Absenteeism 0.00% 0 0,000 0.00%
Total impact on the labour force: -2.30%
Conservative Scenario
i) Contracting flu 30.00% 1 897,600 -2.30%
ii) School closures 15.00% 1 450,000 -1.20%
iii) Absenteeism 20.00% 0 0,000 0.00%
Total impact on the labour force: -3.50%
Severe Scenario
i) Contracting flu 30.00% 1 897,600 -2.30%
ii) School closures 15.00% 1 450,000 -1.20%
iii) Absenteeism 20.00% 1 600,000 -1.50%
Total impact on the labour force: -5.00%
tive and severe cases since we assume a less benign strain of influenza. Further,
we reduce consumption for employees that are absent since we assume this
group undertake behaviour aimed at reducing the likelihood of catching the
disease and reduce travel and social interaction.
The table shows that the baseline scenario produces a -1.2% consumption
shock in 2009Q3 with a 0.8 percent positive, “bounceback” shock in 2009Q4.
The shock is larger in the conservative scenario: -4 percent in 2009Q3 and 1.3
percent in 2009Q4. The severe scenario suggests consumption declines 13%
before increasing by 4 percent in the next quarter. Declines in spending in the
hotels and restaurants category is a key feature of the conservative and severe
scenarios.
As a comparison, the Congressional Budget Office (2006) performs an eval-
uation of the impact on the United States’ GDP for two scenarios: a severe
pandemic of 1918/1919 proportions and a mild pandemic of 1957 and 1968
proportions. The severe pandemic has a gross infection rate of 30 percent and
5Table 2
Impact on consumption
Cons. Sick Absent, Normal Sick Absent, Normal
frac. School School
Closure Closure
Baseline case 09:Q3 09:Q4
Food and beverages 0.2 50% na 100% 138% na 100%
Clothing and footwear 0.06 50% na 100% 138% na 100%
Housing 0.19 100% na 100% 100% na 100%
Household goods 0.16 50% na 100% 138% na 100%
Transport 0.18 0% na 100% 175% na 100%
Hotels and restaurants 0.08 0% na 100% 125% na 100%
Other goods 0.13 50% na 100% 138% na 100%
TOTAL SHOCK: -1.2 % TOTAL SHOCK: 0.8 %
Conservative case 09:Q3 09:Q4
Food and beverages 0.2 20% 50% 100% 160% 115% 100%
Clothing and footwear 0.06 20% 50% 100% 160% 115% 100%
Housing 0.19 100% 100% 100% 100% 100% 100%
Household goods 0.16 20% 50% 100% 160% 138% 100%
Transport 0.18 0% 50% 100% 175% 138% 100%
Hotels and restaurants 0.08 0% 50% 90% 175% 138% 108%
Other goods 0.13 20% 50% 100% 138% 115% 100%
TOTAL SHOCK: -2.7 % TOTAL SHOCK: 1.1 %
Severe case 09:Q3 09:Q4
Food and beverages 0.2 20% 50% 100% 160% 115% 100%
Clothing and footwear 0.06 20% 50% 100% 160% 115% 100%
Housing 0.19 100% 100% 100% 100% 100% 100%
Household goods 0.16 20% 50% 100% 160% 138% 100%
Transport 0.18 0% 50% 100% 175% 138% 100%
Hotels and restaurants 0.08 0% 50% 90% 175% 138% 108%
Other goods 0.13 20% 50% 100% 138% 115% 100%
TOTAL SHOCK: -4.0 % TOTAL SHOCK: 1.3 %
6a case fatality rate of 2.5 percent, whereas the mild pandemic has a gross infec-
tion rate of 20 percent and case fatality rate of 0.1 percent. Under the severe
pandemic an 80 percent reduction (for three months) in the entertainment,
arts, recreation, lodging, and restaurant industries is assumed. Most other in-
dustries suffer a 10 percent reduction in demand except for the government
and education sectors, which have no demand side effect, and health for which
demand increases by 15 percent.
In addition to the labour and consumption shocks listed above, we assume
some impact of influenza on the New Zealand tourism industry in the conser-
vative and severe scenarios only. In particular, we assume that the Northern
hemisphere experiences an influenza epidemic of a similar magnitude to that
suggested in tables 1 and 2. We assume that the shock reduces net exports
of services by 5 percent over 2009Q4 and 2010Q1 (the Northern hemisphere
winter), reducing non-commodity exports by 1.5 percent in each quarter in
the baseline scenario. In the conservative scenario we simply double this effect
and input 3 percent shocks for non-commodity exports in 2009Q4 and 2010Q1.
Finally, we assume that half of this expenditure is returned to New Zealand
in 2010Q2 because a fraction of tourists delay rather than cancel travel plans.
2.4 Macroeconomic projections
With a set of macroeconomic shocks in hand, we use the KITT DSGE model
(see Lees, 2009, for an introduction to the model) to understand the possible
impact of the shocks on the macroeconomy. Confronted with a large labour
supply shock, the model would typically produce a large increase in wages
to encourage labour supply. In this analysis we assume that both workers
and firms recognise the temporary, extraordinary nature of the shock and we
assume no increase in wages for the influenza scenario (by introducing off-
setting shocks).
Figure 1 shows the implied paths of key macroeconomic variables following
the swine flu shock for the preferred baseline case. Consumption immediately
falls as a direct result of the shock. This lowers firms’ investment plans. Net
exports increase initially as a result of a reduction in imports driven by lower
consumption. By the first quarter of 2010, the effect of the influenza attack
on the Northern hemisphere dominates and net exports are actually slightly
negative.
Since the consumption and labour shocks are broadly offsetting, there is only
a small movement in headline consumer price inflation. The impact of the
consumption shock slightly dominates, suggesting some deflationary pressure.
This produces a decline in the policy rate and a concomitant fall in the ex-
7Table 3
Cumulative impact on output
Scenario One quarter One year Three years
Baseline scenario -0.57 -0.62 -0.75
Conservative scenario -1.10 -3.30 -4.05
Severe Scenario -1.52 -8.15 -11.48
change rate.
Figure 2 shows the impact of the more conservative scenario. The fall in con-
sumption is quite dramatic and there is a large fall in investment. The powerful
consumption shock produces a marked improvement in the net export posi-
tion because of a reduction in imports. The effects of the shock to exports
from northern hemisphere tourists reducing spending is clearly visible in the
exports panel. Although the falls in output are much more pronounced than
for the baseline case, the consumption and labour supply shocks are broadly
off-setting with little movement in inflation or the interest rate.
Figure 3 shows the results of the severe scenario. This scenario depicts quite
marked declines in output and key macroeconomic variables, particularly in-
vestment. With less demand for goods, firms reduce production and inputs
to production. Investment falls. However, even under a substantive swine flu
event, it’s not clear that reductions in investment of the magnitude in fig-
ure 3 are realistic and are probably a function of stark simplifying modelling
assumptions regarding how firms produce goods and services.
One key feature of this severe scenario is large cumulated output losses. These
output losses can be compared across the scenarios in table 3.
3 Robustness checks
This paper approaches the question of the magnitude of the impact of swine flu
on the New Zealand economy from a very specific macroeconomic modelling
perspective. There are weaknesses in the model. For example, the model ab-
stracts from the ability of firms to absorb slack in demand by simply allowing
inventories to increase temporarily rather than reducing production substan-
tively. Furthermore, firms may be able to reduce the effective magnitude of the
labour supply shock by simply postponing training and leave, freeing up staff
to replace colleagues absent because of swine flu. To gain some understanding
of the materiality of points such as these, we compare our study with other
studies and historical influenza episodes.
8Fig. 1. Baseline scenario
Consumption Output
0 0
−0.2
−0.5
[%]
[%]
−0.4
−1
−0.6
−1.5 −0.8
2009:3 2010:1 2010:3 2011:1 2009:3 2010:1 2010:3 2011:1
Investment Interest rate
0 0.02
−1 0
−2 −0.02
[%]
[%]
−3 −0.04
−4 −0.06
−5 −0.08
2009:3 2010:1 2010:3 2011:1 2009:3 2010:1 2010:3 2011:1
Headline inflation Exchange rate
0.05 0.1
0.05
0
0
[%]
[%]
−0.05
−0.05
−0.1
−0.1
−0.15 −0.15
2009:3 2010:1 2010:3 2011:1 2009:3 2010:1 2010:3 2011:1
Imports Exports
0 0.04
−0.5
0.03
−1
[%]
[%]
−1.5
9 0.02
−2
−2.5 0.01
2009:3 2010:1 2010:3 2011:1 2009:3 2010:1 2010:3 2011:1Fig. 2. Conservative scenario
Consumption Output
0 −0.2
−0.4
−1
−0.6
[%]
[%]
−0.8
−2
−1
−3 −1.2
2009:3 2010:1 2010:3 2011:1 2009:3 2010:1 2010:3 2011:1
Investment Interest rate
0 −0.06
−2
−0.08
−4
[%]
[%]
−0.1
−6
−0.12
−8
−10 −0.14
2009:3 2010:1 2010:3 2011:1 2009:3 2010:1 2010:3 2011:1
Headline inflation Exchange rate
0.1 0.1
0 0
[%]
[%]
−0.1 −0.1
−0.2 −0.2
−0.3 −0.3
2009:3 2010:1 2010:3 2011:1 2009:3 2010:1 2010:3 2011:1
Imports Exports
0 0.5
−1
0
−2
[%]
[%]
−0.5
−3
10
−1
−4
−5 −1.5
2009:3 2010:1 2010:3 2011:1 2009:3 2010:1 2010:3 2011:1Fig. 3. Severe scenario
Consumption Output
−1 −1
−1.2
−2
−1.4
[%]
[%]
−3
−1.6
−4
−1.8
−5 −2
2009:3 2010:1 2010:3 2011:1 2009:3 2010:1 2010:3 2011:1
Investment Interest rate
0 −0.2
−5 −0.25
[%]
[%]
−10 −0.3
−15 −0.35
−20 −0.4
2009:3 2010:1 2010:3 2011:1 2009:3 2010:1 2010:3 2011:1
Headline inflation Exchange rate
0 0.2
0
−0.2
−0.2
[%]
[%]
−0.4
−0.4
−0.6
−0.6
−0.8 −0.8
2009:3 2010:1 2010:3 2011:1 2009:3 2010:1 2010:3 2011:1
Imports Exports
0 1
−2
0
−4
[%]
[%]
−1
−6
11
−2
−8
−10 −3
2009:3 2010:1 2010:3 2011:1 2009:3 2010:1 2010:3 2011:13.1 Other studies
There are other cross-country studies that we can draw from when considering
the impact of swine flu. Some of the cross-country differences may be large
however and may affect both the attack rate of the influenza (depending on the
density of population and cultural norms) and the macroeconomic impact (for
example, through the impact on net exports). Notwithstanding these points,
table 4 shows the impact of our study in the context of other similar studies.
James and Sargent (2007) produce a very detailed analysis of a range of sce-
narios largely eschewing a macromodelling perspective. Their conclusion is
that the macroeconomic impact of swine flu is likely to limited and produce a
wide range of estimates for three scenarios that range from 0.09 to 1.6 percent
of GDP in the first year following the outbreak.
Keogh-Brown et al. (2009) apply a macroeconomic modelling approach to
a hypothetical influenza outbreak for the UK and suggest a wide range of
estimates of the impact on GDP. The epidemiological assumptions they adopt
are on balance less benign than in this paper (certainly less benign for the
baseline case) but suggest similar macroeconomic findings.
Of particular relevance is the 2006 New Zealand Treasury study by James
Douglas, Kam Szeto and Bob Buckle. Their paper is based on two scenarios:
(i) a severe scenario where the 40 percent of the population is infected with
a fatality rate of 2 percent; and (ii) a mild scenario where 30 percent of the
population is infected with a fatality rate of 0.25 percent. Their scenarios
work from a range of assumptions about declines in demand and increased
absenteeism. Their mild scenario suggests reductions in GDP over a range
between 1-2 percent in year 1 and 1-3 percent over four years. These results
are different to our baseline assumption but in keeping with the different set
of epidemiology assumptions that underpin both our conservative and severe
scenarios.
3.2 Historical episodes
In addition to examining macroeconomic studies, we can look at historical
episodes of influenza outbreaks to help determine possible macroeconomic im-
pacts. Both influenza outbreaks and the impact of the SARS (Severe Acute
Respiratory Syndrome) on Hong Kong and Singapore prove useful cases to
provide a check on the analysis that comes from making microeconomic as-
sumptions and extrapolating to the macroeconomy using a model.
The SARS virus
12Table 4
Comparison with other studies†
Author Country/Region Negative % GDP
James and Sargent 2007 Canada 0.09 to 0.28
0.34 to 0.92
0.4 to 1.1
Jonung and Roeger 2006 EU-25 1.1 to 1.6
McKibbin and Sidorenko 2006 United Kingdom 0.72 to 11.11
Keogh-Brown et al 2009 United Kingdom 0.22 to 4.5
2.52 to 6.05
Douglas, Szeto and Buckle 2006 New Zealand 1-3 (mild case)
10-15 (severe case)
This study New Zealand 0.62 (baseline)
3.3 (conservative)
8 (severe case)
†Other studies relate to influenza or avian flu and not swine flu per se.
‡The assumptions regarding epidemiology and the macroeconomic impact
vary across both “severe” studies.
The SARS virus hit Hong Kong in 2003 infecting around 2,000 people and
killing 300. The main impact of SARS was felt through tourism, declines in
restaurant receipts (see figure 4, from James and Sargent 2007) but only a
limited impact on retail sales. Further, exports of goods were strong through-
out the period. In aggregate, the virus had limited impacts on GDP. Only
minor declines in GDP were recorded relative to the historical volatility in
GDP series.
1918 pandemic
The 1918 pandemic killed about 650,000 people in the US. The epidemic was
characterized by a large spike in a single month in the US. While the toll on
the population was severe, figure 5 (from James and Sargent 2007) shows the
impact on US industrial production was not large. They conclude that while
the structure of the economy may have changed over time, the broad pattern
of data from historical episodes does not support large impacts on output from
outbreaks of influenza.
13Fig. 4. Impact of SARS on the Hong Kong Economy: James and Sargent (2007)
Fig. 5. Impact of 1918 Influenza on US Industrial Production: James and Sargent
(2007)
144 Concluding comments
Estimates based on the latest Ministry of Health information suggest a limited
macroeconomic impact of swine flu on New Zealand. A less benign epidemic
would have more pronounced output and consumption effects although the
inflation and policy effects are small. More pronounced output and consump-
tion effects are largely a function of assumptions made for both the extent of
school closures and prophylactic absenteeism.
Our baseline scenario produces a 0.62 percent decline in output, smaller than
typically suggested in the literature for influenza outbreaks. The ability of
firms to use inventories as a buffer and also reschedule staff activities is not
modelled and may mitigate the impact on output. Furthermore, while our
study is within the broad range of estimates produced by other researchers,
historical episodes of influenza outbreaks do not support large impacts on
the macroeconomy. We believe the impact of swine flu on the New Zealand
macroeconomy is unlikely to be large.
References
Congressional Budget Office, 2006. A potential influenza pandemic: An update
on possible macroeconomic effects and policy issues.
Douglas, J., Szeto, K., Buckle, B., 2006. Impacts of a potential influenza pan-
demic on New Zealand’s macroeconomy. New Zealand Treasury Working
Paper 06/03.
James, S., Sargent, T., December 2007. The economic impact of an influenza
pandemic. Finance Canada Working Paper 2007-04.
Keogh-Brown, M., Wren-Lewis, S., Edmunds, S., Smith, R., 2009. The pos-
sible macroeconomic impact on the UK of an influenza pandemic. Oxford
Economics Department Paper no. 431.
Lees, K., June 2009. Introducing KITT: the Reserve Bank of New Zealand’s
new core macroeconomic model. RBNZ Bulletin.
O’Donovan, B., Zollner, S., 2009. Flu jab: the potential impact of the H1N1 in-
fluenza epidemic on the New Zealand economy. Westpac Institutional Bank,
18 June.
15You can also read
